CostPerGen

Wan 2.2 costs between $0.02 and $0.08 per second of 720p video depending on the provider. Cheapest estimated path: Self-hosted (rented GPU) at $0.02. Prices stamped 2026-08-28; labels below state how each number was obtained.

Video index·Alibaba (open weights)·Updated August 28, 2026

Wan 2.2 pricing

Open-weights video — the only roster model you can price four ways: two aggregators, GPU-time on Replicate, and self-hosting.

How much does Wan 2.2 cost per second?

ProviderPrice /sec (720p)Provider's own unitStatusSource
fal.ai · A14B$0.08$0.08 per secondUSD per output secondDerived2026-08-28source
RunPod Hub$0.06$0.30 per request (5s default)USD per requestDerived2026-08-28source
Replicate$0.021GPU time (~$0.10 per 5s clip)GPU-secondsEstimated2026-08-28source
Self-hosted (rented GPU) (self-host)cheapest$0.02rented GPU timeGPU-hoursEstimated2026-08-28source
How these numbers were computed

What would Wan 2.2 cost you?

Presets:

Model · provider$/sec (720p)TotalStatus
Wan 2.2 · Self-hosted (rented GPU)$0.020—estimated
Wan 2.2 · Replicate$0.021—estimated
Wan 2.2 · RunPod Hub$0.060—derived
Wan 2.2 · fal.ai (A14B)$0.080—derived

Totals use the normalized canonical rate (USD per second of 720p output). Resolution multipliers, minimum charges, and tier effects can change real invoices — check each model page.

Price history

Every future change to a Wan 2.2 price lands in the public change log.

Frequently asked

How much does Wan 2.2 cost per second of video?

Wan 2.2 costs between $0.02 and $0.08 per second of 720p video depending on the provider (stamped 2026-08-28).

What does a 60-second video cost with Wan 2.2?

Between $1.20 and $4.80 for 60 seconds of 720p output, depending on provider.

What is the cheapest way to run Wan 2.2?

The lowest estimated price in the index is $0.02/sec via self-host. Check the label — only live-verified prices are guaranteed current.

Is self-hosting Wan 2.2 cheaper than fal.ai?

At sustained volume, yes — roughly $0.02/sec of output on rented A100 time versus $0.08/sec on fal.ai (both estimates). Below a few hours of output per month, aggregator pricing usually wins once you count setup and idle GPU time.

Compare

Data provenance: FairStack operational cost audit, 2026-03-23 (price-audit-full.json) + launch-spec research pass 2026-08-28. Live re-verification pending (deploy-day). For quality benchmarks (not pricing), seeArtificial Analysis — we index cost, they index quality.